Mapping spin-orbit splitting in strained InGaAs epilayers

Mapping spin-orbit splitting in strained InGaAs epilayers

B. M. Norman, 1 C. J. Trowbridge, 1 J. Stephens, 2 A. C. Gossard, 2 D. D. Awschalom, 2 and V. Sih 1

1 Department of Physics, University of Michigan, Ann Arbor, MI 48109
2 Center for Spintronics and Quantum Computation,
University of California, Santa Barbara, CA 93106
(Dated: June 30, 2010)

Time- and spatially-resolved Faraday rotation spectroscopy is used to measure the magnitude and
direction of the momentum-dependent spin splitting in strained InGaAs epilayers. The epilayers
are lattice-matched to the GaAs substrate and designed to reduce inhomogeneous effects related to
strain relaxation. Measurements of momentum-dependent spin splitting as a function of electron
spin drift velocity along [100], [010], [110] and [110] directions enable separation of isotropic and
anisotropic effective magnetic fields that arise from uniaxial and biaxial strain along (110). We relate
our findings to previous measurements and theoretical predictions of spin splitting for inversion
symmetry breaking in bulk strained semiconductors.

PACS numbers: 71.70.Ej, 71.70.Fk, 72.25.Dc, 72.25.Rb

The polarization and coherent manipulation of elec-
tron spins are important steps towards the realization of
spin-based information processing 1 . While electron spins
can be manipulated by magnetic fields, electrical con-
trol is desirable as it offers the potential for high-speed
manipulation and local gates. Spin-orbit interactions
in semiconductors offer momentum-dependent effective
magnetic fields that can be used for electrical control^!.

Momentum k-dependent spin splittings arise from the
spin-orbit interaction and the breaking of spatial in-
version symmetry. The lack of an inversion center in
zincblende crystal structures results in a bulk inversion
asymmetry (BIA) and the Dresselhaus field 4 . Structural
inversion asymmetry (SI A) along the growth direction in
heterostructures results in the Rashba field 5 . These inter-
nal effective magnetic fields have different k-dependence
and can be distinguished by changing the direction of the
carrier drift momentum. Measurements as a function of
electric field direction have mapped the k-linear Rashba
and Dresselhaus fields in quantum wells 6 .

Strain breaks spatial inversion symmetry, which re-
sults in additional k-dependent spin splittings^] Strain-
induced spin precession and characterization of k-linear
spin splitting have been conducted in lattice-mismatched
heterostructures^ and using a mechanical vise^^.
Measurements on lattice-mismatched heterostructures
observed both BIA- and SIA-type splitting, but the split-
ting did not exhibit a clear trend as a function of mea-
sured strain^. A subsequent theoretical analysis pro-
posed that variations in strain relaxation during growth
could result in different k-linear BIA- type splitting 9 .
Measurements conducted using a mechanical vise found
that uniaxial strain along (110) introduces a SIA-type
splittin^in^.

In this paper we investigate the momentum de-
pendence of strain-induced spin-orbit splittings in
Ino.04Gao.96As epilayers. The InGaAs epilayers are
grown on GaAs substrates, which introduces biaxial com-
pressive strain, and the sample structure was designed
to reduce inhomogeneous effects related to strain relax-

ation. Measurements of internal effective magnetic fields
as a function of electron spin drift velocity along [100],
[010], [110] and [110] enable separation of BIA and SIA-
type spin splittings that arise from biaxial and uniaxial
strain.

For k in the plane perpendicular to the growth direc-
tion [001], the spin-splitting Hamiltonians take the form:

Hd = A (cr x k x ky - (Tykykl) (1)
= a(k y a x - k x a y ) (2)
Hi = D (e zz - e xx ) (a x k x - a y k y ) (3)

H 2 = Cs * xy (k y cr x - k x a y ) (4)

H£> and Hr are the Dresselhaus and Rashba Hamilto-
nians, respectively, while Hi and H2 are two additional
k-linear terms due to strairi^M Here x, and z denote
the [100], [010], and [001] crystal axes, denotes the i th
Pauli matrix, e^- are the components of the strain ten-
sor with e xy = e yx and e xx = e yin and A, a, D, and C3
are material constants. H# and H2 have the same direc-
tion dependence on momentum, while Hi has the same
form as the linear Dresselhaus field for a two-dimensional
system with quantum confinement along [001] 14 . H2 ac-
counts for the spin splitting introduced by uniaxial strain
along (110)ff^^H and Hi has been proposed 9 to explain
the BIA- type splitting in Ref. [10 , despite earlier worlP
that argued that Hi should be small. The directions of
the SIA-type (Hr and H 2 ) and BIA- type (U D and Hi)
fields are shown in Fig. 1(b). For k along [110] and [110],
the SIA and BIA fields are both perpendicular to k and
parallel to each other, while for k along [100] and [010],
the BIA fields are parallel to k while the SIA fields are
perpendicular to k.

In general, H^ is cubic in k while H#, Hi, and H2 are
linear in k. Thus the spin splitting can be described by
combinations of k-linear and k-cubic terms of the form

2

FIG. 1. (Color) (a) Experimental setup. InGaAs channels
and contacts are shown in yellow and orange respectively.
Voltage is applied such that the net electron drift momentum
is either perpendicular to or parallel to the external magnetic
field Bext • The red arrow denotes the laser propagation direc-
tion, (b) Directions of momentum-dependent effective inter-
nal fields due to spin-orbit interactions from structural inver-
sion asymmetry (left) and bulk inversion asymmetry (right).

A = avj + /3vd, where Vd is the in-plane drift velocity
defined by hkd = m*Vd with m* the conduction band
electron effective mass and hkd the drift momentum. Yet
along the [010] axis, k x = and k y = |k^|, so = for
bulk systems. Similarly, = for k along [100]. On the
other hand, for momentum along [110], k x = k y =

and Hp reduces to Hd = ^| (a x — a y ).

The momentum-dependent effective magnetic fields
described above were measured using pump-probe op-
tical techniques. Optical orientation of electron spins
results in an out-of-plane spin polarization, which can
precess about in-plane internal and applied external
magnetic fields. Measurements were performed on Si-
doped Ino.04Gao.96As epilayers (doping concentration n
= 3 x 10 16 cm -3 ). The thickness of the InGaAs layer is
500 nm, grown above a 300 nm growth interrupted GaAs
buffer layer on a (001) GaAs substrate and capped with
100 nm undoped GaAs. X-ray diffraction measurements
show that the InGaAs epilayer is lattice-matched to the
GaAs substrate and exhibits minimal strain relaxation.
Samples were fabricated, consisting of two perpendicular
InGaAs channels [Fig. 1(a)]. Channels were aligned along
[110], [lTO], [100] and [010]. The channels were each 400
/im long and 100 /am wide with 400 x 400 /am 2 ohmic
contacts at each end. Wires were soldered to each of the
contacts and connected to an external voltage source.
The samples were mounted in a liquid helium flow cryo-
stat with measurements made at temperature T = 30
K. Care was taken during sample mounting to minimize
introducing additional strain.

Measurements were conducted using a mode-locked
Ti: Sapphire laser with a repetition rate of 76 MHz tuned
to the band edge of the InGaAs epilayer (A = 848 nm).
The laser was separated into a pump and probe pulse
with a temporal separation controlled by a mechanical
delay line. The pump and probe beams were modulated
by a photoelastic modulator (PEM) and optical chopper
respectively for cascaded lock-in detection. An exter-
nal magnetic field was applied in the plane of the layer
and varied from -40 to 40 mT. The inclusion of a motor

driven steering mirror in the pump path allowed con-
trol of the pump-probe spatial separation on the sample.
Field scans were taken at 5 jam intervals over a 40 jam
range (roughly ±20 jam from the center of the electron
spin packet). By applying an external voltage to the con-
tacts, the electron drift momentum was varied along each
channel. Measurements were performed with the chan-
nels oriented both parallel with and perpendicular to the
applied magnetic field to measure both the magnitude
and direction of the internal magnetic fields.

Faraday rotation (FR) of the probe pulse can be de-
scribed by:

e F = Ae~% cos(^!^|B ext + B int |) (5)

where A is the FR amplitude, g is the electron g factor,
jiB is the Bohr magneton, T| is the inhomogeneous de-
phasing time, At is the pump-probe time delay, and B ext
is the applied external magnetic field. Time-resolved FR
measurements^ find that g = 0.51 and T 2 * = 7.8 ns.
The direction and magnitude of the internal magnetic
field B int , which is proportional to the spin splitting A
= g/i#|Bi nt |, is determined from fits of the FR signal to
Eq. 5 with errors less than 1%. As shown in Fig. 2(a), the
component of Bi nt that is parallel to B ext causes an over-
all shift in the field-dependent signal, and the component
that is perpendicular to B ext changes the magnitude of
the center peak. By (B_l) is defined as the component
of Bi n t that is parallel (perpendicular) to k^, as deter-
mined from field scans based on the channel orientation.
Measurements were taken with k _L B ext and k || B ext
to determine both the sign and magnitude of the spin
splitting. The magnitude of the drift momentum was
varied by changing the potential applied to the channels,
in the range between -2.0V and 2.0V. Scans with varying
pump-probe spatial separations were taken to character-
ize the internal field and drift velocity of the optically
injected spin packet [Fig. 2). The drift velocity Vd of the
spin packet at each voltage is the measured spatial drift
of the spin packet center x c during the pump-probe time
delay At = 13 ns.

For the [010] channel oriented perpendicular to the ex-
ternal magnetic field, values of B^ and By are displayed
for various contact voltages as a function of pump-probe
spatial separation in Fig. 3. The amplitude of the FR is
also fit for each scan to determine the position of the spin
packet center x c at each voltage [Fig. 3(c)]. The spatial
dependence of B int is due to both drift and diffusion of
the spin packet 10 . For the [010] channel, it is expected
that Bi nt nas a linear dependence on k because Hp =
0. For each voltage, a linear fit is used to determine the
value of Bint at x c . These values are fit to linear func-
tions of Vd [Fig. 3(d)] to determine the values of f3± and

Table 1 shows a summary of the measured values of
j3±_ and j3\\ for all four channel directions and both mea-
surement orientations, k || B ext and k _L B ext . All of the
measurements are well described by a linear dependence

3

k Bext

k _L Bext

Channel

H\\

0±

( j q f™,, / n,

o ^xy j > v

Die?? — c^-^/h

J \^-ZZ *~XX f i ' v

0\\

H\\

( j Q f™„ / /7,

v - y o ^xy j > v

J — \^-ZZ *~XX J 1 ' v

[100]

35

-84

-230

47

35

-110

-300

47

[010]

-35

-100

-280

48

-41

-91

-250

56

[110]

-14

-180

70

-20

-150

50

[no]

-120

-94

TABLE I. Measured 0\\ and f3± for each channel and orientation expressed in units of neV ns /im 1 . f3\\ represents the BIA
spin splitting, while f3± represents either the SIA ([100] and [010] channels), SIA plus BIA ([110]) or SIA minus BIA ([110])
spin splitting. Calculated strain parameters C^e X y/fi and D(e zz — e xx )/h are in units of m/s.

FIG. 2. (Color) (a) Faraday rotation 0f as a function of
external magnetic field B ex t for 0.0V (black), 1.0V (red) and
2.0V (green) measured at the center of the spin packet. The
[010] channel was oriented perpendicular to B ex t, so the shift
and decrease in the center peak give the values of B^ and By
respectively, (b)-(d) Faraday rotation as a function of B ex t
and pump-probe separation d for (b) 0.0V, (c) 1.0V, and (d)
2.0V. The data was measured at T = 30 K with a pump-probe
time delay of At = 13 ns.

of Bi n t with k, so P = A/vd- Measurements with k ||
B ext and k _L B ext show reasonable agreement, but with
noticeable discrepancies that are larger than the mea-
surement error, which is typically on the order of 1 — 5
%. Additional strain introduced during sample mounting
and cool-down could possibly account for these discrep-
ancies. For the [100] and [010] channels for which the
SIA- and BIA-type spin splittings can be separated, we
found that the SIA splitting ((3±) is consistently 2 or 3
times larger than the BIA splitting (/3n). For the [010]
channel, the value of /3± can be attributed to H# and
H 2 and j3\\ to Hi, since = for k along [010]. Also,
as expected from the form of Hi [Eq. 3], the BIA-type
splitting (/3||) has similar magnitude but opposite sign for
[100] and [010], which justifies the assumption that e xx

-50 50 -3 3

d (M m ) V d (/xm ns 1 )

FIG. 3. (Color) Internal magnetic fields (a) perpendicular to
and (b) parallel to the channel direction for the [010] channel
at T — 30 K with a pump-probe time delay of At = 13 ns. For
these measurements, the channel was oriented perpendicular
to the applied magnetic field (k J_ B ex t)- Applied voltages are
the same for (a)-(c). Lines in (a) and (b) are linear fits, (c)
Faraday rotation amplitude as a function of pump-probe sep-
aration for various potential differences used to determine the
spin packet center x c . (d) B^ (black) and By (red) evaluated
at x c as a function of spin packet drift velocity. The slope of
the linear fit gives the value of /3/g/HB- Errors in (a)-(c) are
smaller than the data points.

= € yy for this sample.

Although the measurements cannot distinguish be-
tween the contributions of H# and H2, previous mea-
surements indicate that the effect of H 2 is greater
than that of H# for similar bulk epilayers. Measure-
ments^ on unstrained n-doped GaAs epilayers found
f3 < 10 neV ns /im -1 , and that the measured spin split-
ting upon applied strain could be characterized using H 2
and without requiring

From Eqs. 3 and 4 and the measured values of /?, we
calculate the sample parameters C^e xy /h and D(e zz —
e xx)/h assuming no contribution from H^ and H# and
m* = 0.065mo 1T , where mo is the free electron mass.

4

-50 50 -3 3

d (M m ) V d (/xm ns 1 )

FIG. 4. (Color) (a) Internal magnetic field perpendicular to
k measured with k J_ B ex t for the [110] channel at T = 30 K
as a function of pump-probe separation d and with a pump-
probe time delay of At = 13 ns. (b) Internal magnetic field
evaluated at the spin packet center as a function of electron
spin packet velocity. The slope of the linear fit gives the value
of fl/gfiB- Errors in (a) are smaller than the data points.

From the measurements on the [100] and [010] chan-
nels, Cse xy /h has an average value of -270 m/s, and
V(e zz - e xx )/h = 50 m/s. From Ref. [9], C 3 /h = 8 x 10 5
m/s and D/H = (0.5 - 1.5) xlO 4 m/s. From the liter-
ature value of Cs/h and the range of values of Cse xy /h
measured here, e xy = -0.029 — 0.038%. Therefore, the
variation in our measurements indicates that the strain
changes by an amount less than JAe^l = 1 x 10 -4 be-
tween different sample mountings and cool-downs. For
pseudomorphic Ino.04Gao.96As on GaAs 18 , e xx = -0.29%
and e zz = 0.26%, which yields D/ft = 0.9 x 10 4 m/s,
w hich is in the same range as previously reported val-
ue d 9 * 10 *. The data and analysis show that the measured
spin splittings are consistent with Hi and H 2 and that H 2
can have comparable magnitude to Hi , which contra dicts
the argument that Hi is small and can be neglected^ 1 ^.

Unexpectedly, the values of f3±_ for the [110] and [110]
channels do not match the difference and sum, respec-
tively, of the SIA and BIA terms from the other chan-
nels. Therefore, measurements along [110] and [110] may
not be sufficient to accurately characterize Hi and H2.
For [110], = -14 neV ns /im" 1 for k || B ext and -
20 neV ns /im -1 for k _L B ext . These values are both
smaller in magnitude than the average sum (B± + |/3m|)
= -58 neV ns /im" 1 from measurements on the [100] and

[010] channels. Similarly, for [110], f}± = -120 neV ns
/im -1 for k || B ext and -94 neV ns /im -1 for k _L B ext ,
and both values are smaller in magnitude than the aver-
age difference (B± — |/3n|) = -130 neV ns /im -1 . We also
calculate Cse xy /h and D(e zz — e xx )/h assuming that Hi

o< (/?[iio] - P[iio])/ 2 and H 2 oc (Alio] + P[iio])/ 2 and ob-
tain a smaller value for G^e xy /h. The calculated C^e xy /h
and T)(e zz — e xx )/h values are included in Table 1. We
hypothesize that higher order terms, including the addi-
tional k-cubic Dresselhaus term (Hd) for [110] and [110],
could account for this discrepancy.

We also observe unexpected behavior in the measured
slope of Bi n t as a function of pump-probe spatial sepa-
ration and voltage for the [110] channel. The slope of
Bi n t at each voltage is due to diffusion and the distribu-
tion of drift velocity of the spin packet from the finite
spot size ( 30 x 10 /im) of the pump and probe beams^.
Electrons at the leading edge of the packet have a higher
average velocity than those in the back and thus we ex-
pect a non-zero and positive slope. However, as shown in
Fig. 4(a), for scans in the voltage range 1.0 - 2.0V for the
[110] sample, surprisingly, a negative slope is observed.
This was found from repeated measurements on different
spots along the channel and is therefore not thought to
be a result of any interface reflection effects. The mea-
surements of j3 do not seem to be affected, as the value
for Bint for the center of the spin packet at each volt-
age still has a linear dependence on spin drift velocity,
as shown in Fig. 4(b). Measurements for both configura-
tions k || B ext and k _L B ext show reasonable agreement,
with values for /3± of -14 and -20 neV ns /im -1 , respec-
tively. The smaller spin splitting could possibly explain
why this behavior is only apparent in the [110] sample.
Larger values of (3 could overwhelm this effect in other
channels.

This material is based upon work supported by the
National Science Foundation under Grants No. ECCS-
0844908 and DMR-0801388 and the Horace H. Rackham
School of Graduate Studies. Sample fabrication was per-
formed in the Lurie Nanofabrication Facility, part of the
NSF funded NNIN network.

1 S. A. Wolf, D. D. Awschalom, R. A. Buhrman, J. M.
Daughton, S. von Molnar, M. L. Roukes, A. Y. Chtchelka-
nova, and D. M. Treger, Science 294, 1488 (2001).

2 S. Datta and B. Das, Appl. Phys. Lett. 56, 665 (1990).

3 D. Awschalom and N. Samarth, Physics 2, 50 (2009).

4 G. Dresselhaus, Phys. Rev. 100, 580 (1955).

5 Y. A. Bychkov and E. I. Rashba, J. Phys. C 17, 6039
(1984).

6 L. Meier, G. Salis, I. Shorubalko, E. Gini, S. Schon, and
K. Ensslin, Nature Physics 3, 650 (2007).

7 G. E. Pikus and A. N. Titkov, in Optical Orientation,
edited by F. Meier and B. P. Zakharchenya (Elsevier Sci-

ence Publishers, Amsterdam, 1984).

8 G. C. La Rocca, N. Kim, and S. Rodriguez, Phys. Rev. B
38, 7595 (1988).

9 B. A. Bernevig and S.-C. Zhang, Phys. Rev. B. 72, 115204
(2005).

10 Y. Kato, R. C. Myers, A. C. Gossard and D. D.
Awschalom, Nature 427, 50 (2004).

11 S. A. Crooker and D. L. Smith, Phys. Rev. Lett. 94, 236601
(2005).

12 M. Beck, C. Metzner, S. Malzer and G. H. Dohler, Euro-
phys. Lett. 75, 597 (2006).

13 V. Sih, H. Knotz, J. Stephens, V. R. Horowitz, A. C. Gos-

5

sard, and D. D. Awschalom, Phys. Rev. B 73, 241316(R)
(2006).

R. Winkler, Spin- Orbit Coupling Effects in Two-
Dimensional Electron and Hole Systems, Springer Tracks
in Modern Physics Vol. 191 (Springer, Berlin, 2003).
M. Hruska, S. Kos, S. A. Crooker, A. Saxena and D. L.

Smith, Phys. Rev. B 73, 075306 (2006).

16 S. A. Crooker, D. D. Awschalom, J. J. Baumberg, F. Flack
and N. Samarth, Phys. Rev. B 56, 7574 (1997).

17 I. Vurgaftman, J. R. Meyer, and L. R. Ram-Mohan, J.
Appl. Phys. 89, 5815 (2001).

18 S. C. Jain, M. Willander, and H. Maes, Semicond. Sci.
Technol. 11, 641 (1996).